Data Tech 2026

Agentic AI and prompt engineering are dominating the headlines, but in the industrial world, a wrong prediction doesn’t just show a bad ad or a wrong purchase recommendation—it breaks a machine. Deploying AI in physical settings involving IoT, sensor data, and engineering constraints is fundamentally different from consumer tech, and in this space, physics still outranks prompts. If data science teams and business leaders don’t have a shared understanding of these unique industrial challenges, project timelines can easily slip, and quality can suffer; conversely, when technical teams miss the commercial context, they risk building elegant models that solve the wrong problems.

This talk serves as a "two-way street" translation guide to bridge the gap between technical execution and business strategy. We will explore 5 timeless realities, with examples, required to make industrial AI work—ranging from the integration of domain knowledge and physics-based feature engineering, to the critical necessity of industrial-grade MLOps and strict performance benchmarking.